An inset groundwater-flow model to evaluate the effects of layering configuration on model calibration and assess managed aquifer recharge near Shellmound, Mississippi

Abstract

The U.S. Geological Survey has developed a high-resolution inset groundwater-flow model in the Mississippi Delta as part of an interdisciplinary collaboration coordinated by the Mississippi Alluvial Plain project to provide a tool that stakeholders can use to support water-resource management decisions. Groundwater withdrawals from the Mississippi River Valley alluvial (MRVA) aquifer have been vital to support agricultural production in the region, but substantial groundwater-level declines near Shellmound, Mississippi, have caused concerns for long-term sustainability of the aquifer. To better understand the subsurface and try to mitigate the long-term groundwater-level declines, stakeholders have undertaken actions including a Groundwater Transfer and Injection Pilot (GTIP) project using a riverbank filtration-based managed aquifer recharge approach. The pilot project consisted of extracting groundwater near the Tallahatchie River and reinjecting it into the aquifer 3 kilometers west where water levels have substantially declined. A high-resolution airborne electromagnetic (AEM) survey was also completed to collect electrical resistivity data to support the GTIP project and the development of the groundwater model. The inset groundwater-flow model was developed to (1) integrate the AEM data into the optimal layering configuration of the MRVA aquifer that the available observation data can support through calibration, and (2) assess the potential effect of the GTIP project on the groundwater levels. The AEM data were processed into three different layering configurations leading to the development of model A (18 layers), model B (16 layers), and model C (8 layers), all at a 100- x 100-meter cell spatial resolution using the U.S. Geological Survey modular finite-difference flow model 6 code with Newton-Raphson formulation. The model development process integrated recent advances in modeling, such as the incorporation of AEM data, the use of outputs from the soil-water-balance (SWB) model, and the Aquaculture and Irrigation Water-Use Model, and was facilitated by robust automation using the open-source python packages Modflow-setup and SFRmaker. Using Parameter Estimation ++ Iterative Ensemble Smoother, the three numerical groundwater-flow models (models A, B, and C) were calibrated against a set of observations, which included aquifer groundwater levels, streamflows, stream stage, and aquifer transmissivity. Results indicate that the detailed representation of MRVA aquifer layers in model A produced the best calibrated model by history matching, and the integration of data representing surficial connectivity played a key role in improving groundwater recharge and enhancing the ability of the model to match groundwater levels in the cone of depression. A forecast model simulated the managed aquifer recharge approach, and the results indicated that, given average irrigation and recharge conditions (2010–15), the GTIP project has the potential to induce groundwater-level increases of as much as 3 meters around the injection site, but a sustained increase would require repetition in subsequent years of water transfer at 2022 rates or above.